Estimating the dimensionality of learning: The model for decomposed change

Abstract

Psychometrics traditionally has assumed that the changes occurring during learning have the same dimensionality as the targeted construct. This approach is akin to understanding learning as climbing a mountain: for each ability a student can increase their ability, decrease it, or not change it. However, an alternative approach has recently been suggested – learning score analysis. This analysis is grounded in the assumption that students “slide through” the curriculum rather than “climb it”: they forget something to learn something. This approach suggests that positive and negative dynamics can occur at the same time. This suggestion is also supported by the evidence from cognitive psychology that learning and forgetting are separate cognitive processes. Hence, we contrast the traditional approach to conceptualizing growth and change to the alternative approach. We use IRT-based modeling to compare the results from both approaches, and show that the alternative approach provides more insight into learning.